A Novel Role for FERM Domain-Containing Protein 3 in CKD

Abstract

Key Points:

1. We have identified a transcriptional signature of 93 genes associated with CKD severity and progression.

2. Protein 4.1, ezrin, radixin, moesin domain-containing protein 3 gene expression is reduced in the context of more severe kidney disease and in individuals who go on to develop progressive disease.

3. Protein 4.1, ezrin, radixin, moesin domain-containing protein 3 interacts with proteins of the cell cytoskeleton and cell-cell junctions in proximal tubule epithelial cells.

Background: Currently, there are limited methods to link disease severity and risk of disease progression in CKD. To better understand this potential relationship, we interrogated the renal transcriptomic profile of individuals with CKD with measures of CKD severity and identified protein 4.1, ezrin, radixin, moesin-domain containing protein 3 (FRMD3) as a candidate gene for follow-up study.

Methods: RNA-sequencing was used to profile the transcriptome of CKD biopsies from the North Dublin Renal BioBank, the results of which were correlated with clinical parameters. The potential function of FRMD3 was explored by interrogating the FRMD3 interactome and assessing the effect of lentiviral mediated FRMD3 knock down on human renal proximal tubule epithelial cells by assessing cell viability, metabolic activity, and structural markers.

Results: We identified a subset of 93 genes which are significantly correlated with eGFR and percentage tubulointerstitial fibrosis at time of biopsy and with CKD progression 5 years postbiopsy. These results were validated against transcriptomic data from an external cohort of 432 nephrectomy samples. One of the top-ranking genes from this subset, FRMD3, has previously been associated with the risk of developing diabetic kidney disease. Interrogating the interactome of FRMD3 in tubule epithelial cells revealed interactions with cytoskeletal components of cell-cell junctions. Knockdown of FRMD3 expression in tubule epithelial cells resulted in increased proapoptotic activity within the cells, as well as dysregulation of E-Cadherin.

Conclusions: We have identified a panel of kidney-specific transcripts correlated with severity and progression of kidney disease, and from this, we have identified a possible role for FRMD3 in tubule cell structure and health.

Graphical abstract

Figure 1

Transcriptomic analysis of human kidney biopsy samples links gene expression and disease severity. (A) Heatmap of gene expression in human kidney biopsy samples from the NDRBB and eGFR, as well as %TIF, with 935 genes correlating with both in our cohort (n=24). Of these, 359 genes showed higher expression in the setting of more severe disease (B), being eGFR Neg and %TIF Pos, and 574 genes showed higher expression in the setting of less severe disease (C), being eGFR Pos and %TIF Neg. eGFR Neg, negatively correlated with eGFR; eGFR Pos, positively correlated with eGFR; NDRBB, North Dublin Renal Biobank; %TIF Neg, negatively correlated with %TIF; %TIF, percentage tubulointerstitial fibrosis; %TIF Pos, positively correlated with %TIF.

Figure 2

Differential expression analysis of progressive versus stable CKD highlights key dysregulated processes and displays overlap with markers of disease severity. Gene set enrichment analysis of genes differentially expressed within patients with progressive CKD versus Stable CKD was performed using the Enrichr platform and the Reactome and Go Biological Processes databases with the top 50 processes from each dataset by combined score used for visualization and analysis. Genes upregulated in the progressive CKD group (A) were enriched for immune system-related terms, while those downregulated in the progressive CKD group (B) were enriched for cell and mitochondrial metabolic processes. (C) Comparing genes which correlate with eGFR and %TIF in either direction with genes which are differentially expressed between the progressive CKD group and the stable CKD group reveals 93 genes which were common across all three analyses. ER, endoplasmic reticulum; GPI, glycosylphosphatidylinositol.

Figure 3

FRMD3 expression levels are negatively associated with disease in the human kidney. FRMD3 expression by RNA-seq inversely correlated with (A) %TIF (r=−0.73; P = 4.817×10⁻⁵) and positively with (B) eGFR (r=0.62; P = 0.001175) at time of biopsy in a cohort of patient kidney biopsies (n=24). (C) FRMD3 expression in kidney cell types by single-nuclear RNA-seq. Using a publicly available single-nuclear RNA-seq dataset of patient kidney biopsies from the Kidney Precision Medicine Project, FRMD3 expression was found to be expressed in many cell types within the kidney including tubule and glomerular cell types. (D) FRMD3 levels in the kidney precision medicine project single-nuclear RNA-sequencing dataset from healthy reference versus patients with CKD shows a reduction of FRMD3 levels averaged across all clusters in CKD versus healthy reference samples (average log₂fold-Change −0.34, P_adj = 2.43×10⁻²⁸⁸) using the Wilcoxon rank-sum test with Bonferroni correction in Seurat. ****P < 0.0001. (E) FRMD3 levels in the proximal tubule cluster of the kidney precision medicine project single-nuclear RNA-sequencing dataset from healthy reference versus CKD patients shows a reduction of FRMD3 levels in CKD versus healthy reference samples (average log₂fold-Change=−0.53, P_adj = 3.44×10⁻¹³⁹) using the Wilcoxon rank-sum test with Bonferroni correction in Seurat. ****P < 0.0001. (F) Using a publicly available single-nuclear RNA-seq dataset from the Humphries laboratory, accessed at http://humphreyslab.com/SingleCell/, FRMD3 was again found to be expressed in many cell types within the kidney, with lower expression observed throughout in early diabetic nephropathy patients versus control samples. For box and whisker plots within violin plots (C–E), the box ranges from the 25th percentile to the 75th percentile, with whiskers extending 1.5 times the IQR past those points. The internal line represents the median. ATL, ascending thick limb; CD-ICA, collecting duct intercalating cell A; CD-ICB, collecting duct intercalating cell B; CD-PC, collecting duct principal cell; CNT, connecting tubule; DCT/CT, distal convoluted tubule/connecting tubule; DTL, distal thick limb; EC, endothelial cell; ENDO, endothelial cell; FIB, fibroblast; FRMD3; protein 4.1, ezrin, radixin, moesin domain-containing protein 3; IC, intercalating cell; IMM, immune cells; LEUK, leukocytes; LOH, loop of Henle; MES, mesenchymal cells; NEU, Schwann cell/neural; PapE, papillary tip epithelial cell; PC, principal cell; PCT, proximal convoluted tubule; PEC, parietal epithelial cells; PEC, parietal epithelial cell; POD, podocyte; PODO, podocyte; PT, proximal tubule; RNA-seq, RNA-sequencing; TAL, thick ascending limb; VSM/P, vascular smooth muscle/pericyte.

Figure 4

FRMD3 is localized to cell-cell contacts in human proximal tubule cell monolayers. V5-tagged FRMD3 overexpressing HK-2 cells were used to visualize the location of FRMD3 within the cell by laser scanning confocal microscopy z-stack (A). At low density, FRMD3 appears diffusely spread throughout the membrane of the cell, while at high density FRMD3 appears to be enriched at cell-cell contacts (yellow arrows). The main image at each density (high density: 63× magnification, low density: 40× magnification, scale bars 20 µm) represents a single optical section of the sample; the yellow crosshairs represent the area taken as orthogonal slices displayed to the right (YZ) and bottom (XZ) of the main images. The red boxes around the XZ orthogonal sections mark the area enlarged to the right of both sets of images. Proteomic analysis of FRMD3-V5 interactors in these cells reveals interactions between FRMD3 and proteins linked with maintenance of the F-actin cytoskeleton (purple), including RHO-family GTPases like CDC42 and RAC1, as well as Integrin B1 (B, adapted from Ingenuity pathway analysis, Qiagen). HK-2, human proximal tubule epithelial. RHO, RAS homolog.

Figure 5

Knockdown of FRMD3 in human proximal tubule cells disrupts cell junction protein levels. To recapitulate the lower levels of FRMD3 seen within patients with CKD and diabetes nephritis, a HK-2 was used to generate stable shRNA knockdowns against FRMD3 or a nontargeting control shRNA. (A) The FRMD3 knockdown cells showed a 0.8228±0.030-fold reduction in FRMD3 mRNA levels compared with the scrambled non-targeting control shRNA expressing cells (****P < 0.0001, delta delta CT method, B-actin housekeeping control). (B) To investigate the role of FRMD3 within epithelial cell junctions, FRMD3 knockdown and scrambled control shRNA expressing HK-2 cells were treated with 2.5 ng/ml TGFβ1 or vehicle for 48 hours. After this treatment, we performed Western blot against E-cadherin, a key marker of epithelial cell integrity (densitometric analysis on n=3 replicates normalized to α-tubulin expression) (C) representative western blot for (B). FRMD3 knockdown resulted in a basal 2.4-fold increase in E-cadherin levels (P_adj = 0.0001). TGFβ1 treatment resulted in an almost 90% decrease in the amount of E-cadherin protein within the scrambled control shRNA expressing cells (88.8% reduction, P_adj = 0.0031). (D) Proteomic analysis of FRMD3 interactors in FRMD3-V5 overexpressing HK-2 cells also revealed interactions with components of the adherens cell-cell junction (purple), including the scaffold protein IQGAP1, delta1 Catenin (CTNND1), and with components of the tubulin cytoskeleton (TUBA4A, TUBA4B, TUBB, TUBB3, TUBB6, TUBB8) (adapted from ingenuity pathway analysis, Qiagen). shSCR, short hairpin scrambled control RNA; shRNA, short hairpin RNA.

Figure 6

Knockdown of FRMD3 in human proximal tubule cells results in decreased cell viability. HK-2 expressing either a FRMD3-targeting shRNA or a control nontargeting shRNA (shSCR) were used to investigate metabolic function and viability in cells with reduced FRMD3 expression. MTT and Resazurin assays were used to assess NAD(P)H-dependent metabolic function within these cells. Knockdown of FRMD3 resulted in a 39.29%±5.244% (P = 0.0017) reduction in NAD(P)H metabolic activity compared with the scrambled nontargeting shRNA expressing cells (A), agreeing with the resazurin assay showing a reduction of 30.24%±8.433% (P = 0.0230) (B). Values were normalized to total mitochondrial content using Mitotracker Green. (C) A synthetic fluorescent substrate (synthetic fuorogenic substrate for caspase 3 (Acetylated)) was used to assess the activity of proapoptotic effector caspases (caspase 3 and caspase 7) in the FRMD3 knockdown cells, showing an increase in activity of 58.91%±8.935% (P = 0.0006) compared with the scrambled nontargeting control shRNA expressing cells. (D) This disruption to NAD(P)H-dependent metabolism was observed in the absence of disruption to oxygen-dependent cell metabolism as measured by the mito stress test assay using the Seahorse XF pro instrument. This implies the observed results were not because of large-scale mitochondrial dysfunction. For all, n=3 experiments. MTT, methylthialazole tetrazolium; NAD(P)H, dihydrogennicotinamide adenine dinucleotide phosphate; OCR, oxygen consumption rate.